摘要：In image classification, one of the most significant elements is the broad range of data, particularly when supervised learning is used for the classification of images. Several reports are produced by experts who are experienced in their specialty. Even though a huge amount of medical data is complicated and has an expensive procedure which needs teamwork between the scientists and the clinics. The issue is mostly tried to be solved with the utilization of conventional techniques of data augmentation, creating a few adjustments to images of dataset for example rotating, zooming, cropping and size. In this proposed work, the modern technique of data augmentation is shown which is known as DC-GAN i.e., Deep Convolutional Generative Adversial Network. This is a procedure to produce artificial medical images. Moreover, for the improvement of DR, we will take the help of the classification model that is resnet50 for the eye related classification. The suggested technique is shown on the APTOS-Blindness dataset. First, the current online data augmentation methods are used and the production of artificial images of retina take place with the help of DCGAN. Now, we use the method of classification for both the techniques. In the end, after the training of method takes place by utilizing the original and artificial clinical images, the outcome shows- the suggested model identifies all the stages of DR unlike the present methods and achieve accuracy of 98.66.
